Welcome to the October 27, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.
ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version.
The online version now has a button at the top labeled "Show Headlines."
|
|
|
AI Model Fundamentally Cracks CAPTCHAs
National Public Radio
Merrit Kennedy
October 26, 2017
Researchers at the artificial intelligence company Vicarious say they have developed a computer model that can solve Completely Automated Public Turing Tests to Tell Computers and Humans Apart (CAPTCHAs). Vicarious co-founder Dileep George says the model cracks a CAPTCHA's defenses by parsing the text the test presents more effectively than earlier deep-learning models and with less training. George cites the model's use of a Recursive Cortical Network, which is better able to reason about what it is seeing with less training than previous models. He says the network's training entails constructing internal models of the letters it is exposed to, and when a new image is presented it attempts to explain it, trying to define all of the pixels of that new image in terms of the characters it has previously seen. George notes the long-term objective of such experiments "is to build intelligence that works like the human brain. CAPTCHAs were just a natural test for us."
|
Falling Behind
Inside Higher Ed
Jeremy Bauer-Wolf
October 27, 2017
Job growth in the computing field is far outstripping the supply of students earning bachelor's degrees in computer science and similar disciplines, according to a new report from the U.S. National Academies of Science, Engineering, and Medicine. "Strains on educational institutions are significant," the report says. "There is a growing sense of an impending crisis in many universities." The academies recommend colleges and universities consider channeling more resources into computer science departments to address professors' mounting workloads, while also teaching courses creatively with more emphasis on technology for "high-quality instruction." The report also recommends that institutions attempt to boost student interest in computer science, and deliberately seek to attract and retain more women and minority learners. "The broad opportunities in computing in both the labor market and for enabling a host of intellectual pursuits will continue to be drivers of increasing enrollments in undergraduate computer science, from both majors and non-majors," the report says.
|
Research Team Led by NUS Scientists Breaks New Ground in Memory Technology
National University of Singapore
October 24, 2017
Researchers at the National University of Singapore (NUS) say they have developed a thin, organic film that supports 1 million more read-write cycles and consumes 1,000 times less power than commercial flash memories. In addition, the technology can store and process data for 1 trillion cycles and has the potential to shrink even smaller than its current size of 60 square nanometers. "Our work shifts the paradigm on how the industry has traditionally viewed organic electronics, and expands the application of such technologies into new territories," says NUS professor T. Venky Venkatesan. The researchers say they created the technology by fabricating a novel organic resistive memory device that outperforms commercial flash memory in terms of endurance, energy efficiency, and cost. The device utilizes a transition metal complex that can serve as the basis for engineering new materials by bringing together different variations in the molecules, adding active functions, and using specific counter ions.
|
An AI Has Learned How to Pick a Single Voice Out of a Crowd
New Scientist
Richard Gray
October 24, 2017
Researchers at the Mitsubishi Electric Research Laboratory in Cambridge, MA, have developed an artificial intelligence (AI) system that can separate the voices of multiple speakers in real time, which they say could give automatic speech-recognition technology a major boost. The system uses a machine-learning technique called "deep clustering" to identify unique features in the voiceprint of multiple speakers. It then groups the distinct features from each speaker's voice together, enabling it to separate multiple voices and reconstruct what each person was saying. Mitsubishi's Niels Meinke says the system can separate and reconstruct the speech of two people speaking into a single microphone with up to 90-percent accuracy. However, Meinke notes with three speakers the accuracy falls to 80 percent, while in preliminary tests the system was able to simultaneously separate the voices of up to five people. "The system could be used to separate speech in a range of products including lifts (elevators), air-conditioning units, and household products," Meinke says.
|
Let Your Car Tell You What It Needs
MIT News
David L. Chandler
October 25, 2017
Researchers at the Massachusetts Institute of Technology (MIT) are developing software that measures automotive sounds and vibrations and communicates any service requirements and other diagnostic information to drivers via a smartphone. To develop and evaluate the diagnostic systems, the team tested data from various cars, often renting the vehicles, creating conditions they wanted to be able to diagnose, and then reverting the autos to normal. MIT researcher Joshua Siegel says many of the diagnostics stem from the use of machine-learning processes to compare numerous recordings of sounds and vibrations from well-tuned cars with similar vehicles that have a specific problem. Siegel notes these systems can then filter out even highly subtle differences. For example, he says algorithms for identifying wheel balance problems are more accurate at detecting imbalances than expert drivers from a major automaker. Siegel says a prototype smartphone app incorporating these diagnostic tools should be ready for field-testing in about six months.
|
IBM Scientists Say Radical New 'In-Memory' Computing Architecture Will Speed Up Computers by 200 Times
KurzweilAI.net
October 25, 2017
IBM Research scientists on Tuesday announced the first "in-memory computing" architecture, which is expected to speed up computers 200-fold and facilitate ultra-dense, low-power, massively parallel computing systems. The architecture involves using one device for information storage and processing as a replacement for the standard "von Neumann" architecture. "This is an important step forward in our research of the physics of [artificial intelligence], which explores new hardware materials, devices, and architectures," says IBM's Evangelos Eleftheriou. "As the CMOS [complementary metal-oxide semiconductor] scaling laws break down because of technological limits, a radical departure from the processor-memory dichotomy is needed to circumvent the limitations of today's computers." IBM's Abu Sebastian notes their work conclusively demonstrates how they can leverage "the physics of these memory devices to also perform a rather high-level computational primitive. The result of the computation is also stored in the memory devices, and in this sense the concept is loosely inspired by how the brain computes."
|
Researchers Unveil Tool to Debug 'Black Box' Deep-Learning Algorithms
Columbia News
October 24, 2017
Researchers at Columbia and Lehigh universities have developed DeepXplore, a debugging tool for error-checking deep-learning neural networks by feeding them confusing real-world inputs to reveal infrequent cases of flawed reasoning. "You can think of our testing process as reverse-engineering the learning process to understand its logic," says Columbia's Suman Jana. "This gives you some visibility into what the system is doing and where it's going wrong." DeepXplore detects more bugs than random or adversarial testing by using the neural network to generate test images likely to induce clashing decisions by neuron clusters, and employing optimization to prompt as many conflicting decisions as possible while maximizing activated neurons. Applying DeepXplore to 15 real-world neural networks uncovered thousands of previously overlooked bugs, and brought overall accuracy up to 99 percent in some networks. The researchers will present DeepXplore next week at the ACM Symposium on Operating Systems Principles (SOSP 2017) in Shanghai, China.
|
Radboud Wins 2017 Minerva Informatics Equality Award
Informatics Europe
October 24, 2017
The Institute for Computing and Information Sciences (iCIS) at Radboud University in the Netherlands was named the winner of the second edition of the Minerva Informatics Equality Award for its contribution to promoting women's careers in informatics. This year's award focused on initiatives encouraging the transition of female Ph.D. and postdoctoral researchers into faculty positions. Radboud professor Lejla Batina and researcher Bernadette Smelik received the prize at a special ceremony held during the European Computer Science Summit at the University of Lisbon in Portugal. The iCIS at Radboud University has had a tradition of monitoring gender equality issues since 2001, taking actions such as active recruitment of female researchers, multidisciplinary and international recruitment, support of female staff and researchers by their managers, guidelines for promotions to senior positions, and a mentoring program for women. The award is sponsored by Google and carries a prize that will be used for further work on promoting gender equality.
|
Old Phones Get New Life in High-Powered Computer Servers
Princeton University
Susan DeSantis
October 24, 2017
Researchers at Princeton University have demonstrated that old smartphones can be used to build computer servers at much lower cost than high-end servers. "You can get decommissioned smartphones at low cost because used phones are inexpensive and no one else wants them," says Princeton professor David Wentzlaff. He also notes lower operational costs partly stem from the smartphone batteries' energy efficiency. The researchers say a server could be composed of 80 to 100 smartphones placed in a mobile device cage in a specially designed metal box. Wentzlaff says the phones will be wired together and linked by an Ethernet connection to combine the computing power of the individual chips, and the metal box addresses concerns about possible fires sparked by lithium batteries. "In most of the scenarios, our cost of ownership is cheaper when compared to an industrial server that has the same computational, memory, and storage capacity," notes Princeton's Mohammad Shahrad.
|
Caltech Opens a Drone Lab, With Big Ideas to Improve How Robots Work With Humans
Los Angeles Times
Rosanna Xia
October 24, 2017
The California Institute of Technology (Caltech) on Tuesday officially opened its Center for Autonomous Systems and Technologies (CAST), where researchers will test concepts for improving robots and drones' ability to think and react independently so human-robot interaction and collaboration can improve. CAST director Morteza Gharib says the facility's researchers already are undertaking ambitious "moonshot" challenges, including a high-speed self-flying ambulance and a swarm of self-configuring drones that generate large-scale maps or images. CAST contains an indoor arena where various environmental conditions can be replicated for drone training and testing. Lorraine Fesq at the Jet Propulsion Laboratory of the National Aeronautics and Space Administration says students at CAST will need training to understand all aspects of creating a robotics system. "We hire them, and we have them work over the summer and during the school year to help us...on these projects," she notes. Caltech's Jacob Reher sees the center as a place in which robot development can enter new dimensions via multidisciplinary collaboration.
|
Using Robots in Patient Rehabilitation
American Associates, Ben-Gurion University of the Negev
Andrew Lavin
October 23, 2017
Researchers at Ben-Gurion University of the Negev (BGU) in Israel conducted a study to test user preferences when engaging with a robot on a joint movement task as an initial step toward developing an interactive movement rehabilitation protocol. "Our research shows that the type of motions that the robot makes when interacting with humans makes a difference in how satisfied the person is with the interaction," says BGU's Shelly Levy-Tzedek. Twenty-two college-age participants played a leader-follower mirror game with a robotic arm; when the arm was leading, it executed movements that were either sharp or smooth. Levy-Tzedek says the study supported three conclusions--robotic movement primes human movement, there is no clear-cut preference for leading or following the machine, and a human preference for smooth movements exists. "Thus, determining the elements in the interaction that make users more motivated to continue is important in designing future robots that will interact with humans on a daily basis," Levy-Tzedek says.
|
Secure Payment Without Leaving a Trace
Karlsruhe Institute of Technology
Monika Landgraf
October 17, 2017
Researchers at the Karlsruhe Institute of Technology (KIT) in Germany have developed a secure and anonymous system for making online payments. They say the system is basically an "electronic purse" that works anonymously, while also preventing misuse. The "black-box accumulation plus" (BBA+) protocol developed by the team transfers all necessary account data to the card or smartphone used and guarantees its confidentiality with the help of cryptographic methods. In addition, the team notes BBA+ offers security guarantees for the operator of the bonus or payment system, including a correct account balance that is mathematically constructed so the identity of the user is disclosed as soon as an attempt is made to pay with a manipulated account. "Our new protocol guarantees privacy and security for customers during offline operation as well," notes KIT's Andy Rupp. The researchers will present the system next week at the ACM Conference on Computer and Communications Security (CCS 2017) in Dallas, TX.
|
|
